Celebrity AI Voices: Is It Legal to Clone a Famous Voice?

Using a celebrity's voice in text-to-speech without their permission is, in most places, both unethical and a legal risk. A person's voice is treated as part of their identity, and courts have ruled against advertisers who used sound-alikes even when no real recording was copied. The only safe route is written consent from the person (or their estate), with clear limits on what the voice may say.

This guide explains why, walks through the real cases that shaped the rules, and shows what to do if you want a memorable voice without borrowing someone else's.

Why is a celebrity's voice protected at all?

A famous voice is valuable for the same reason a famous face is: people recognize it, trust it and associate it with a specific person. When a synthetic voice sounds like that person, listeners assume they said it, or at least agreed to it. That assumption is what the law protects.

Three ideas do most of the work.

Right of publicity

In the United States, most states recognize a "right of publicity": the right to control commercial use of your name, likeness and other identifying traits. California's statute (Civil Code section 3344) names "voice" explicitly. Other countries reach a similar result through privacy law, personality rights, or false-endorsement and passing-off claims.

The key word is commercial. Using a recognizable voice to sell a product, promote a channel or make an app more attractive is exactly the use these laws target.

False endorsement

If a voice makes people think a celebrity backs your product, you may also face a false-endorsement claim, separate from publicity rights. The harm is to the public as well as the celebrity: listeners are misled about who is speaking.

Copyright protects recordings, not the voice itself

A common misunderstanding is that a voice can be copyrighted. Generally it can't. Copyright protects specific recordings, songs and scripts. That cuts both ways: training a clone on someone's copyrighted recordings raises its own questions, but the voice as a sound is protected mainly through publicity and endorsement law, not copyright.

What are the real cases that set the rules?

None of these are hypothetical. They are the examples lawyers still cite.

Case or event	What happened	Why it matters
Midler v. Ford (US, 1988)	After Bette Midler declined to sing in a car commercial, the agency hired a backup singer to imitate her. The court found for Midler.	Imitating a distinctive voice to sell a product can violate publicity rights even with no original recording used.
Waits v. Frito-Lay (US, 1992)	A snack ad used a singer imitating Tom Waits's gravelly voice. Waits won, including on a false-endorsement claim.	Sound-alikes can imply endorsement, which is a separate wrong.
"Heart on My Sleeve" (2023)	A track using AI-imitated vocals of two well-known artists spread online and was pulled from major streaming services.	Platforms act quickly on AI vocal imitations, regardless of how the track was made.
New Hampshire robocall (January 2024)	Voters received calls with an AI-generated imitation of President Biden's voice telling them not to vote in the primary.	Led the US Federal Communications Commission to rule in February 2024 that AI-generated voices count as "artificial" voices under robocall rules.
Tennessee ELVIS Act (2024)	Tennessee expanded its publicity law to cover voice explicitly, including AI simulations of it.	Lawmakers are now writing AI voice imitation directly into statutes.

The pattern is consistent: the closer the imitation, and the more commercial or deceptive the use, the worse the outcome for whoever made it.

Is a parody or "sound-alike" any safer?

Sometimes, but less than people think.

• Parody and commentary get more room in many legal systems, especially when the use is obviously a joke about the person and not an ad. The protection depends heavily on context and jurisdiction.
• Labelling it "AI" or "impression" helps with honesty but doesn't by itself fix a publicity-rights problem. Midler and Waits were sound-alikes, and the advertisers lost anyway.
• Political content is a special risk. Faked candidate voices can break election and robocall rules, and several jurisdictions have passed or proposed laws aimed at deceptive election deepfakes.
• Disclosure duties are spreading. The EU AI Act requires that audio which imitates a real person in a way that could pass as genuine be disclosed as artificially generated or manipulated.

This article is general information, not legal advice. If you're planning a commercial project around a real person's voice, talk to a lawyer in the country where you'll publish.

What are the ethical problems, even when it's legal?

Law sets the floor. Ethics asks a broader question: would the person be comfortable hearing this?

Consent is the whole question

A voice is personal in a way few other assets are. Hearing yourself say words you never said is unsettling even when the content is harmless. Without consent, a clone takes something from the person, whatever your intentions are.

Misuse is easy and hard to undo

A clone that can say anything will eventually be asked to say something harmful: a fake apology, a fake endorsement, a fake emergency call. Once audio spreads, corrections rarely catch up with it.

Listeners deserve to know

People make decisions based on who they think is speaking. A synthetic voice that passes as a real person, without telling the listener, misleads them even if the subject never finds out.

Voice actors earn their living with their voices

Cloning a professional voice without a licence replaces paid work. Many voice actors now negotiate specific clauses about AI training and synthetic reuse in their contracts, and those terms should be respected.

How do you use a real person's voice the right way?

If you have a legitimate reason to use a specific person's voice, such as your own founder, a host who has agreed, or a narrator you've hired, this is the process.

1. Get written, informed consent. The person should know it's a synthetic voice, what it will be used for, where it will be published and for how long.
2. Define the scope. List what the voice may say (product tutorials, for example) and what it may not (politics, health claims, anything about third parties).
3. Agree on payment and term. A one-off fee, a per-use fee or a revenue share, with an end date and the right to withdraw.
4. Keep the source recording and the agreement together. If anyone asks later, you can show where the voice came from.
5. Disclose synthetic speech where listeners could otherwise be misled.
6. Review periodically. Rules are changing quickly, and the person's comfort may change too.

In Leelo, voice cloning works from a short recording and is built around consent: you clone your own voice, or the voice of someone who has agreed to it. The step-by-step guide to cloning your voice covers how to record a clean sample.

How do you get a memorable voice without imitating anyone?

Most people reach for a celebrity voice because they want a voice with character: authority, warmth, grit, humour. You can get those qualities without borrowing an identity.

Design an original voice from a description

Leelo's voice design creates a new voice from a written description, for example "a calm, low male voice in his sixties with a slight rasp, unhurried, like a late-night radio host." The result belongs to your project and doesn't belong to anyone else. See how to design a voice.

Cast from a library built for the purpose

The voice library has hundreds of voices in more than 80 languages. Listen for timbre and pace instead of searching for a famous match. Leelo doesn't offer celebrity voices.

Direct the performance

A lot of what makes a famous voice recognizable is delivery, not timbre: long pauses, a particular rhythm, emphasis on unexpected words. You can shape all of that line by line with speed, pitch, pauses and style. The guide on directing the delivery shows how.

Worked example: an ad that wanted "a famous movie-trailer voice"

A small studio wants the classic trailer feel for a product launch video. Instead of imitating a known trailer narrator:

• They design a voice from "deep, resonant, slow, dramatic male narrator."
• They write the script in short, heavy phrases: "This autumn. One app. Every language."
• They slow the speed slightly and add long pauses between phrases.
• They put a low boom from the sound effects lane under each phrase.

The result has the genre feel the brief asked for, with no single identifiable person behind it, and they can use it commercially on a paid plan without a licence negotiation. Plan details are on the pricing page.

For fictional characters (think Yoda or Darth Vader), the issues overlap but are not identical. We cover those separately in why iconic character voices are so hard to copy.

Frequently asked questions

Is it illegal to make an AI voice that sounds like a celebrity?

It depends on the use and the country, but commercial use without permission is risky almost everywhere. US courts have ruled against sound-alike ads since the 1980s, and newer laws, such as Tennessee's ELVIS Act, address AI voice imitation directly. Private, non-commercial experiments carry less risk, but publishing them still can create problems.

Can I use a celebrity voice if I say it's AI-generated?

A disclaimer helps listeners, and in some places disclosure is required, but it doesn't grant you rights to the person's voice. The sound-alike advertisers in the Midler and Waits cases never claimed to be the singers, and they still lost. Consent is what makes the use legitimate.

Can I clone the voice of someone who has died?

Many jurisdictions protect publicity rights after death, often for decades, and those rights pass to the estate. You would need the estate's permission for commercial use. Ethically, consider how the person and their family would feel about new words in their voice.

Does Leelo offer celebrity voices?

No. Leelo offers a library of hundreds of ready-made voices, voice design from a description, and voice cloning for your own voice or the voice of someone who has given consent. If you want a distinctive sound, design an original voice and direct its delivery.

What should a voice consent agreement include?

At minimum: who the voice belongs to, that it will be synthesized, the permitted uses and forbidden uses, where it will be published, payment, duration, and how consent can be withdrawn. Keep the agreement with the original recording used to create the voice.